基于邻域回归的医学图像超分辨率重建

为了提高医学图像的分辨率，提出一种基于内部样例的邻域回归的超分辨率方法。首先，把输入的低分辨率图像当做高分辨率图像去构造基于自身实例的内部图像训练集，不再依赖外部训练集；然后，把高分辨率重建分成高频重建和低频重建，用邻域回归方法重建图像高频细节部分，用双三次插值方法重建低频部分；最后，用迭代组合的方法联合高频分量和低频分量来获得最终输出的高分辨率图像。实验结果表明，该方法性能优于传统的超分辨率重建算法，重建出的医学图像视觉效果更真实。

Medical image super-resolution reconstruction based on neighborhood regression

This paper proposed a super-resolution method based on neighborhood regression of internal examples, in order to improve the resolution of medical images. First, it treated the input low-resolution image as a high-resolution image to construct an internal image training set based on its own instance, no longer dependent on the external training set. Then, it divided the high-resolution reconstruction into high-frequency reconstruction and low-frequency reconstruction, neighborhood regression method reconstructed the high-frequency detail part of the image, and bicubic interpolation method reconstructed the low-frequency part. Finally, iterative combination method combined the high-frequency component and the low-frequency component to get a high-resolution image of the final output. The experimental results show that the proposed method outperforms the traditional super-resolution reconstruction algorithm, and the reconstructed medical image visual effect is more realistic.